Changes in Nikkei 225 Composite Stocks- Trading Simulation ... · stocks (deletions) from the...

1

© 2005, Security Analysts Association of Japan

Feature Article: Asset-Backed Securities

Changes in Nikkei 225 Composite Stocks- Trading Simulation and Implication

Katsuhiko Okada Managing Director

IZ Management Co. Ltd (Chartered Member of the Security Analysts Association of Japan)

Table of Contents

Introduction

１． Previous Studies and the Theory of Index Change ２． Research Design ３． Empirical Findings

A. Stock price behavior around the announcement date B. Existing hypotheses and interpretation

４． Trading Simulation A. Two types of trading strategy B. Simulation results

C. Alternative hypotheses Conclusion

Some 39 index composite changes in the period between 1991 and 2002 were analyzed in the traditional

event study framework. Statistically significant abnormal returns on both added and deleted stocks were

observed not only on the day of announcement but also other days around the announcement date. Trading

simulation based on typically employed trading strategies by proprietary dealers was also conducted. The

simulation results reveal that the same trading strategies paid off repeatedly over the 12-year sample period.

Considering the fact that index composite change is a recurrent event, one may argue that there is a case where

arbitrage activity may not necessarily rule out mispricing.

Katsuhiko Okada: B.A. in Law Kwansei Gakuin University, 1985 /MBA Washington University John M. Olin School of Business, 1990./

M.A, in Business Administration Kobe University, 2004./ Associate Options Trader (1990-1992) Morgan Stanley & Co. Ltd Tokyo,

Functional Vice President in Derivatives (1993-1996 )UBS Securities Tokyo. Chief Investment Officer (1997-2002) PHI KOM Management

Pte Ltd Singapore (Hedge Fund Management), Managing Director (2003-Present) IZ Management Co. Ltd (Investment Advisory)

Tables and Figures are tentatively shown in another PDF file (Only titles of them are translated into English).

2


Introduction

In financial economic theory, there is a long-held assumption that stocks have perfect substitutes and perfect

elasticity of demand. Therefore, any supply or demand shocks that have no informational content should have little

impact on stock price. Since stock price should be identical to the discounted future stream of cash flow that a firm is

expected to generate, demand or supply size is an irrelevant issue for pricing. Theoretically, if a firm’s shares are traded

at an equilibrium price, a large buy order from a brokerage house is supposed to have little influence on the price level.

The price should only move when the fundamental value of the firm changes. This may seem counter-intuitive,

however the argument is supported by the existence of arbitrageurs. If a stock is priced more than its underlying

fundamental value, a rational arbitrageur comes in and sells the stock and buys a similar stock whose fundamental

value is fairly priced. The condition for arbitrageurs to freely buy (sell) undervalued (overvalued) stocks in the market is

that a perfect substitute is always available. Under this condition, the demand curve of the stock would be flat. Scholes

(1972） investigated large block order trades on NYSE. He found that stock price movement in the face of large sell

(buy) orders is limited. His findings indicate that stock price reaction to a sell order is neutral when the seller is a trust

fund or estate.1 He insists that stock price does not respond to non-information. There are quite a few empirical studies

on whether price changes occur in the absence of new information. They focused on block trades, equity issues, and

stock splits. If stock price is independent from the size of demand as financial theory suggests, the predicted share price

movement in response to a change in the stocks comprising an index is neutral. Since index change is an event that has

little to do with a firm’s fundamental value, a rational arbitrageur will come in and correct it should the price move upon

the news. However, in previous studies, empirical evidence is not in line with this prediction. In the majority of index

changes in the US, additions go up in price upon announcement while deletions lose value. There are several

hypotheses to explain these price moves depending on post event abnormal returns but little evidence that stock price is

immune.

The main purpose of this paper is to clarify the price movement of both added stocks (additions) and deleted

stocks (deletions) from the Nikkei 225 based on the equivalent methodology of US studies. There are two

broadly-quoted equity market indexes in Japan. One is the capital-weighted average of all listed stocks on the Tokyo

Stock Exchange 1st Section (TOPIX) and the other is the price-weighted index of 225 representative stocks (Nikkei

225). The focus of index changes here is the price-weighted Nikkei 225. The Nikkei 225 is unique in its calculation

method. While most actively traded index futures in industrialized countries are derived from capital-weighted indexes

such as the S&P 500, the Nikkei 225 is a simple average of 225 issues divided by the divisor published by the Nikkei

Shimbun. Therefore, this is the first paper that analyzes price-weighted index changes. Secondly, this paper investigates

whether arbitrage activity functions to eliminate mispricing in the Japanese market. For this objective, a typical

1 Trust funds and estates use block trades when they mechanically adjust their portfolios. Therefore their sell (buy) orders do not contain any information about firms’ fundamental value.

3


proprietary dealer’s position on the announcement of index change is simulated. The result of this simulation shows that

arbitrage activity does not necessarily eliminate mispricing as presupposed in the theory. The fact that the same trading

strategy paid off over the decade cast doubt on arbitrageurs’ price correction ability over a short horizon.

Watakori (2000) and Saito and Onishi (2001) analyzed the Nikkei 225 change in April 2000. Since this

particular event had a substantial impact on the overall stock market, it attracted widespread attention. Hanaeda and

Serita (2003) employed a similar methodology used in most US studies and found that the demand curve of Japanese

stocks is downward sloping. As for minor indices, Serita (1996) used the Nikkei 300 and Liu (2000) conducted a

similar study using the Nikkei 500. Both identified that the event had a significant impact both on stock price and

trading volume.

This paper analyzes all index changes since 1991 using the traditional event study framework. Previous study

by Hanaeda and Serita (2003) dealt with only one event even though the stocks that were switched numbered 40.

Therefore, any conclusion drawn from the analysis is not free from sampling bias. In particular, the change in April

2000 is considered to be a rather special case, and one may argue that it may not be representative of normal market

response to index change. Studying the 39 index changes over the 12-year period overcomes such a problem.

Simulation based on arbitrageurs’ position taking revealed an interesting pattern in stock prices. The same

trading strategy generated a lucrative return over the period. This demonstrates that arbitrageurs failed to eliminate

mispricing in this particular setting. Despite the fact that index changes are widely known and repeated events,

mispricing persisted in our sample period.

In Section 2, prior literature and hypotheses regarding index changes are introduced. Section 3 describes the

research design and Section 4 empirical findings. Trading simulation is conducted in Section 5 and a conclusion given

in Section 6.

1. Previous Studies and the Theory of Index Change

Previous studies in the US reveal that index changes affect the performance of stocks concerned to a

considerable degree. When changes are announced, prices of newly added stocks go up by a substantial margin, while

deleted stocks decrease in value. As for the rationale behind this there are mainly four hypotheses discussed in previous

studies. However, researchers have shown competing evidence in a variety of event samples, thus no consensus has

been reached as yet. The four hypotheses that are commonly quoted in the literature will be discussed.

The imperfect substitute hypothesis2 attributes the observed market response to the lack of similar assets

2The imperfect substitute hypothesis describes the situation where arbitrageurs cannot freely trade the misaligned stocks due to lack of close substitutes in the market.

4


available in the securities market. It argues that arbitrageurs’ activity is limited because they cannot sell (buy) the

overvalued additions (undervalued deletions) due to the lack of substitutes. The efficient market hypothesis (EMH)

assumes that arbitrageurs would counter-sell (short) the overbought additions and buy the stocks of firms whose

business content is similar to the ones they are short. After a while, the value of those two groups of stocks should

converge, bringing about risk-free profit to arbitrageurs. EMH insists, therefore, regardless of the size of demand at the

time of composite change, that stock prices remain the same as demand is information free. In the real world, however,

similar stocks that can be substituted for arbitrageurs’ short positions are not easily available. When additions face a

large number of temporary bidders, the increase in demand will lead to higher stock prices. The stocks remain unduly

expensive because the lack of substitutes makes it prohibitively difficult for arbitrageurs to short them. Thus,

incompleteness of the securities market undermines the fair pricing of the stocks concerned. The imperfect substitute

hypothesis is alternatively called the downward sloping demand curve hypothesis (DSDC) as information free demand

moves stock prices. Shleifer (1986) found additions to the S&P500 rise approximately 3% on average. He supported

DSDC by indicating that the rise was not temporary. Wurgler and Zhuravskaya (2002) investigated S&P 500 Index

changes between 1976 and 1996 and found the adopted stocks rose 3.5% on average. They also supported the DSDC

hypothesis for the same reason. They looked into the arbitrage risk of stocks in the market and found only 25% had

close substitutes. The lack of close substitutes in the market makes it difficult for arbitrageurs to collectively participate

should the stock price be derailed. Hanaeda and Serita (2003) investigated the relationship between the cumulative

abnormal return of additions and arbitrage risk. They found a positive relationship between them implying that the lack

of close substitutes makes stock price vulnerable to a demand shock.

The price pressure hypothesis, on the other hand, describes price action as a result of market maker behavior.

When additions to an index are announced, it creates a temporary positive demand shock. There are a large number of

bidders for the stocks and the market has to absorb them. Market makers take the opposite side of the trade. They will

unwind the position gradually after the effective date. However, they have to carry the position for several days and are

subject to price risk during that period. To compensate for that risk, market makers demand a premium and thus the

prices of stocks concerned rise beyond fair value.

Harris & Gurel (1986) investigated S&P500 composite changes between 1978 and 1983. They found the

additions rose both in price and volume but that the temporary impact faded in two weeks. They concluded the rise was

a simple reaction to temporary price pressure.

Given that all the available information is reflected in the stock price, the price rise of additions is due to

non-public information contained in the event itself. The announcement of index additions conveys some information

about the fundamentals of the stocks. For instance, additions to an index may increase awareness of the stocks among

securities analysts. Being included in analysts’ coverage will eventually increase a firm’s brand awareness among

5


investors. Jain (1987) investigated the sub-indexes of the S&P 500 and found that stock prices rise when additions take

place not only with respect to the major indexes that institutional investors follow but also minor sub-indexes. Beneish

and Gardner (1995) found that additions to the Dow Jones Industrial Average exhibited little change in price and

volume while deleted stocks declined in price and volume to a statistically significant extent. Both of these findings are

consistent with the information hypothesis. Simply being one of the members of the index composite has some value,

which is priced in the market.

2. Research Design

An event study framework is employed for analysis of composite changes between 1991 and 2002. The stock index

futures market started in 1988 and the change in 1991 was the first major one since the futures market became liquid.

There were a total of 39 index changes during this period. In every case, announcement was made after the market had

closed and publicized in the morning edition of the Nikkei Shimbun the following day. There is a one- to 30-day interval

until any new composite of the index becomes effective. Investors tracking the index adjust their portfolios during this

period. This paper looks at two different event days, one when the change is publicly announced which is defined as the

announcement day (AD), and the other when the actual change takes place and which is defined as the change day

(CD).

Table 1 shows the 39 cases from 1991. The total number of stocks added/deleted was 178. These 39 index

changes include periodic review by the Nikkei Shimbun as well as the bankruptcy and merger of constituent firms,

which occasionally brings the number of stocks comprising the Nikkei 225 to fewer than 225 for a few days. Normally,

there are two to 10 days between AD and CD, but bankruptcy necessitates immediate change. For example, when

Nikkatsu went bankrupt on July 1st, 1993, newly added Iseki Machinery started trading as a constituent of the Nikkei

225 on July 2nd.

6


No AD CD Deletions Additions No AD CD Deletions Additions

1 1991 9 25 ～1991 10 1 Taito Kumagai Constr. 14 2000 6.23. ～ 2000. 7.3. Tonen Corp ShiseidoKatakura Sumitomo H.I.

Teikoku Textile Topy Industry 15 2000. 9.8. ～ 2000 9.22. IBJ Yokohama B.Matsuzakaya Tomen Corp DKB Toyo Trust

Shochiku Nissho Iwai Corp Fuji Bank Shinko Sec.Toho Sankyu

2 1992 9 18 ～1992 9 24 Japan Stainless Aoki Constr. 16 2000. 9.8. ～ 2000 9.26. KDD Alps Elec.

3 1992 9 25 ～1992 10 1 Godo Shusei Hazama Const.Daito Boseki Ninebea 17 2000. 9.8. ～ 2000. 10.2. Tekken Constr. MizuhoTakashimaya Seika Ind. Japan Se. Fin. Secom

4 1993 3.26. ～1993. 4. 1. Sanyo National Pulp Iseki Mach. 18 2001 3. 9. ～ 2001.3.23. Japan Paper JAL5 1993 7. 1. ～1993. 7. 2. Nikkatsu Shionogi Pharm.

6 1995 9 22 ～1995 10 2 Japan Wool Marui 19 2001.3.9 ～ 2001.3.27. Sakura Bank Takashimaya7 1996 3 19 ～1996 3 25 Bank of Tokyo Chubu Elec Bank of T.M. Credit Saison8 1996 9 17 ～1996 9 24 Honshu Paper Sanwa Bank Sanwa Bank Yamato Trans.9 1997 9 17 ～1997 9 24 Mitsui Toatsu Toyo Rubber Tokai Bank --

10 1998 9 17 ～1998 9 24 Japan Cement Asahi Bank Mitsu Trust --Showa Marine Tran. KDD Toyo Trust --

11 1999 3 16 ～1999 3 25 Mitsubishi Oil NTT DataNavix Line Clarion 20 2001.3.9 ～ 2001.3.30 Japan Unibac

12 2000 3 21 ～2000 3 28 Mitsui Trust Daiwa Sec.21 2001.3.9 ～ 2001.4.3. MTFG

13 2000 4.14. ～2000 4.24. Nichiro JT UFJ Holdings.Mitsui Mining Kao

Sumitomo Coal Daiichi Pharm 22 2001.9.11 ～ 2001.9.25 Sumitomo Marine JR WestJapan Tensaito Eisai

Honen Termo 23 2001.9.11 ～ 2001.10.1. Iseki Mach. Sekisui HouseFuji Textile TDK Keihin Railway Fujisawa Pharm

Toyo Rayon Mistumi ElecRasa Industry Matsushita Comm 24 2001.11.27. ～ 2001.11.28 Niigata Iron W. Sumitomo Real E.Japan Carbide Adventest 25 2001.11.26. ～ 2001.12.5. Daiwa B. --

Japan Chemical Casio 26 2001.12.6. ～ 2001.12.7. Aoki Const. DaikinJapan Synthetics Fanuc 27 2002.11.26. ～ 2001.12.12 -- DB Holdings

Asahi Denka Kyocera 28 2002.2.15. ～ 2002.2.25 Asahi B. --Nihon Yushi Taiyo Yuden 29 2002.2.22 ～ 2002.2.27. -- Chiba B.

Toyo Rubber Matsushita Denko 30 2002.3.3. ～ 2002.3.4. Sato Kogyo --Japan Carbon MMC 31 2002.3.3. ～ 2002.3.6. -- Japan Comsis

Noritake Fuji Heavy 32 2002.3.19. ～ 2002.3.26. Tokyo Marine --Shinagawa Ref. Tokyo Electron 33 2002.3.19. ～ 2002.4.2. -- Mikea Holdings

Japan Metal Seven-Eleven 34 2002.9.5. ～ 2002.9.6. Fujita Const. --Nihon Yakin Ito-Yokado 35 2002.9.5. ～ 2002.9.11. -- MT HoldingsNihon Denko Jusco

Mitsubishi Steel IBJ 36 2002.9.5. ～ 2002.9.19. Japan Energy Trend Micro

Shimira Daiwa B. Kwasaki Steel OlympusShowa Cable Tokai B. NKK IsetanTokyo Seiko Shizuoka B.

Japan Piston Ring Sumitomo Trust 37 2002.9.5. ～ 2002.9.25. Matsushita Comm --

Seika Ind. Yasuda Trust JAL --Iwatani Ind. Sumitomo Marine

Maruzen JR East 38 2002.9.5. ～ 2002.9.25. -- Shin Nikko

Sankyu DDI -- JFEMistui Warehouse NTT-Docomo

39 2002.9.5 ～ 2002.10.2. Kyokuto Boeki CSKTobishima Const. Japan Air Sys.

Table 1: Index Change Events between 1991 and 2002

(Note) AD refers to the date when the annoucement of index change is published in the media. Actual composite chage does not occur until theeffective date denoted as "CD".There are total of 39 index change events since 1991.

7


The Nikkei 225 is a price-weighted average of its constituent stocks.3 It simply adds the prices of the 225 stocks

and divides by a divisor calculated by the Nikkei Shinbun. When an index change is announced, the index is calculated

in this manner until the new composite of stocks becomes effective on the CD. After the CD, the new member firms’

shares are added and divided by the new divisor. The index has to maintain its price sequence over time. Therefore, the

new divisor is calculated to guarantee the sequence before and after the CD. The following equation has to be satisfied.

The sum of the pre-change 225 stocks / the pre-change divisor = the sum of the post-change 225 stocks / the new

divisor.

Index tracking investors who wish to minimize risk due to composite change would like to sell stocks to be

deleted and buy stocks to be added one day prior to the CD. To be specific, if they could swap the deletions and

additions at the closing price, they would bear no risk.

This event study uses the market model4. The estimation of beta is based on the 221 days of a time series return

of the market and the stocks. Thus, between 250 days and 30 days prior to the AD is the estimation period. TOPIX is

used as a proxy for market return. Table 2 shows the beta of both additions and deletions. For example, the beta value of

the deleted portfolio in event 1 is 1.0899. This is estimated from the time series data of the simple average of the three

deleted firms’ return and the market return. The R-square value is 0.3587 and the t-value rejects the null hypothesis that

the beta is zero at the 1% significance level. Events 13 and 14, however, fail to reject the null hypothesis.5 This

describes one aspect of the peculiarity of the composite change in April 2000. Witness the contrast between deletions

and additions. The composite change in April 2000 was peculiar in its size, content, and timing. The deletions generally

belong to the so-called “old economy” and additions to the “new economy”. While Watakori (2001) criticizes the way

the change was conducted, it should be noted that the April 2000 event was distinctly different from the others in

various aspects. Table 2 is a simple manifestation that the choice of the deletions and additions was unusually biased.

Using the beta in Table 2, abnormal return is calculated as follows:

τττ β miii RRAR^

−= （1）

, where

τiAR is the abnormal return of portfolio i at day τ

i

^β is the estimated ß of portfolio i

3 Japan Railway, Japan Tobacco, and NTT DoCoMo are calculated by converting them to 50 yen face value. 4 The market model assumes the time series model as itmtiit RR εβ += . This study estimated beta using (-250,

-30）estimation period. The intercept and the risk rate used in CAPM are disregarded. 5 The market model was used to estimate abnormal return for these two events.

8


τmR is TOPIX return at dayτ

τiR is portfolio i ’s return at dayτ

1 1.0899 0.3587 11.5873 1.4754 0.7843 29.5383

2 1.0480 0.3258 10.7686 1.7381 0.6113 19.4283

3 1.5786 0.4071 12.8376 1.3644 0.7030 23.8362

4 1.3456 0.4355 13.6354 1.4992 0.4910 15.2484

5 0.9472 0.2245 8.3707 1.0034 0.2829 9.7703

6 1.5267 0.4625 14.4601 1.0060 0.3036 10.2930

7 1.3208 0.5535 17.3933 0.3208 0.1735 7.1564

8 1.0383 0.2242 8.3982 1.3955 0.4313 13.6037

9 1.2840 0.3498 11.3858 0.8543 0.1975 7.7011

10 1.7772 0.3865 12.3483 1.5559 0.5038 15.6745

11 1.3463 0.3930 12.5678 0.9708 0.1421 6.3582

12 0.9518 0.0629 4.0233 1.3457 0.3075 10.3458

13 0.0687 0.0013 0.5714 0.9632 0.7440 26.5191

14 0.1936 0.0085 1.4441 0.1373 0.0173 2.0702

15 0.7450 0.1156 5.6469 0.8571 0.2245 8.4043

16 1.4777 0.1934 7.6493 0.6525 0.0551 3.7736

17 0.2390 0.0173 2.0713 1.3989 0.1633 6.9014

18 0.4675 0.0519 3.6544 0.5212 0.0486 3.5316

19 0.7068 0.2333 8.6175 0.6026 0.1913 7.5983

22 0.2918 0.0753 4.4585 0.4356 0.0626 4.0353

23 0.6442 0.2237 8.3845 0.3554 0.0759 4.4755

24 1.1370 0.2258 8.4522 0.7472 0.1965 7.7412

25 1.2599 0.2687 9.4690 Deletion only

26 1.3934 0.1476 6.5138 0.9338 0.1924 7.6401

28 1.3974 0.2307 8.5361 Deletion only

29 Addition only 0.6493 0.1625 6.8669

31 Addition only 0.5692 0.0759 4.4589

32 0.8154 0.3063 10.3580 Deletion only

36 1.0302 0.3640 11.8182 0.8809 0.4873 15.2276

37 1.2232 0.5120 15.9998 Deletion only

39 0.8000 0.3251 10.8404 1.2572 0.3487 11.4308

Note: 1. Portfolio returns are calculated by simple arithmatic average of composite stocks.

　2. Portfolio betas are estimated based on the 200 day estimation period, which is AR-230 to AR-31.

　3. Exiting portfolio of event No. 13 & 14 were unable to reject their null hypothesis that beta is zero.

　　4. I used market model to estimate the beta

results

ßAdjusted R-

sqt-value

EventNumber

ßAdjustedR-square

t-value

Table 2: Beta Estimates For the Index Changes

Exiting Portfolio Entering Poirfolio

9


The event window is defined as 10 days prior to the AD and 25 days after the CD. The abnormal return is calculated

during this period. Statistical significance is measured using the t-test. t-values are computed as follows:

i

iAR

ARt

i στ

τ ˆ= （2）

, where

τiARt is the t value of portfolio i

iσ̂ is portfolio i ’s standard deviation of its time series data for 221 days. When the composite change greatly affects

stock price on a single day, it is detected by τiARt . If the change affects stock price for a few days, it is reflected in the

cumulative abnormal return (CAR). The statistical significance of CAR is measured by τiCARt . t-value of event i is

denoted as,

i

iCAR

d

CARt

i στ

τ ˆ= (3)

, where

τiCAR is the cumulative abnormal return of portfolio i at day τ .

d is the number of days abnormal returns are accumulated.

iσ̂ is portfolio i ’s standard deviation of its time series data for 221 days.

When stocks are deleted from the index, they are excluded from our sample since they decrease in value due to

bankruptcy rather than being deleted. Other stocks outside the sample include Millea Holdings, Mitsui Banking and

Trust Holdings, Shin Nikko Holdings, JEF Holdings and Japan Air System Holdings. There is not sufficient pre-event

price history available for any of these to estimate beta.

Mean abnormal return (MAR) and mean cumulative abnormal return (MCAR) were also calculated. MAR is

the average abnormal return of i portfolio across all events. Likewise, MCAR is the average cumulative abnormal

return of i portfolio across all events.

∑=

=39

1

1

iiAR

mMAR ττ (4)

, where

τMAR is the average abnormal return of i portfolio across 39 events. m denotes the number of events available for

analysis.6 Table 3 shows the MCAR from 10 days prior to the AD. These figures are computed from

∑−=

=τ

τ10t

tMARMCAR (5)

, where τMCAR is the ten-day cumulative figure of tMAR

6 Some events consist of only deletions while others only additions. Some have only one day between AD and CD and others have 16. Therefore, one has to count the number of events available in a given day for MAR calculation.

10


3. Empirical Findings

Section A demonstrates abnormal return distribution surrounding the announcement day, and section B explores

a possible explanation of the price pattern in relation to hypotheses found in previous literature on the subject.

A. Stock price behavior around the announcement date

Table 3 shows MAR and MCAR around the AD and CD. The event window period is defined as （-10, 10）,

-10 being 10 days prior to the announcement date and 10 being 10 days after it. The third column indicates the number

of events. Note that the number of events in the event window drops dramatically for deletions after the CD. This is due

to the fact that the majority of deletions are stocks that disappeared from the market. Many of them were merged or

went bankrupt. They remained tradable after announcement but not over the effective date. Eventually, only seven

events remained over the CD. The fourth column shows the standard deviation of the 221-day time series data of the

addition (deletion) portfolio and the last column is the cumulative MAR from 10 days prior to the event day.

As described in the introduction, additions to (deletions from) the index itself do not convey any fundamental

information. In other words, such non-informational events should not affect stock price. However, the statistically

significant reaction of both the additions and deletions portfolio on AD+1 should be noted. The temporary demand

increase (supply increase) in the stocks in the additions (deletions) portfolio affects the value of the stocks to a

statistically significant extent.

Suppose the composite change event has some informational value, its content should be reflected on the day of

the announcement. However, one can also observe statistically significant MARs on other days as well. For instance,

on AD+1, AD+2, and AD+3, MARs are significantly negative for deletions and positive for additions. It is interesting

that stock prices fail to fully discount the information upon receipt.7

7 Serita （1996） explains this phenomenon using non-uniformity of information.

11


Days relative Deletions Days relative Deletions

to event MAR # of event t-value MCAR to event MAR # of event t-value MCAR

-10 -0.21% 27 -0.426 -0.21% -10 0.29% 27 0.579 0.29%

-9 0.40% 27 0.807 0.19% -9 -0.37% 27 -0.738 -0.08%

-8 -0.43% 27 -0.867 -0.24% -8 -0.44% 27 -0.886 -0.52%

-7 -0.16% 27 -0.322 -0.41% -7 -0.06% 27 -0.127 -0.59%

-6 0.01% 27 0.017 -0.40% -6 -0.51% 27 -1.011 -1.09%

-5 -0.07% 27 -0.142 -0.47% -5 -1.26% 27 -2.518 *** -2.36%

-4 -1.00% 27 -1.992 ** -1.47% -4 -1.75% 27 -3.499 *** -4.11%

-3 -0.33% 27 -0.655 -1.79% -3 -2.52% 27 -5.033 *** -6.64%

-2 -0.44% 27 -0.884 -2.24% -2 -2.32% 27 -4.617 *** -8.95%

-1 -0.15% 27 -0.304 -2.39% -1 -3.61% 27 -7.209 *** -12.56%

AD -0.33% 27 -0.660 -2.72% CD 4.81% 7 6.033 *** -7.75%

1 -4.50% 26 -8.805 *** -7.22% 1 1.26% 7 1.579 -6.49%

2 -1.43% 26 -2.800 *** -8.65% 2 -0.28% 7 -0.357 -6.78%

3 -1.03% 24 -1.905 ** -9.68% 3 -1.44% 7 -1.805 ** -8.22%

4 -0.78% 19 -1.263 -10.46% 4 2.03% 7 2.542 *** -6.19%

5 -0.85% 18 -1.432 -11.31% 5 -0.44% 7 -0.555 -6.63%

6 -1.02% 15 -1.670 * -12.33% 6 0.55% 7 0.693 -6.08%

7 -1.13% 14 -1.890 * -13.46% 7 -0.35% 7 -0.434 -6.42%

8 0.21% 14 0.349 -13.25% 8 -0.34% 7 -0.425 -6.76%

9 -1.10% 10 -1.386 -14.35% 9 0.08% 7 0.106 -6.68%

10 -1.16% 10 -1.465 -15.51% 10 -0.29% 7 -0.362 -6.97%

Days relative Additions Days relative Additions

to event MAR # of event t-value MCAR to event MAR # of event t-value MCAR

-10 0.15% 27 0.376 0.15% -10 0.52% 27 1.271 0.52%

-9 -0.12% 27 -0.294 0.03% -9 -0.71% 27 -1.723 * -0.19%

-8 0.24% 27 0.580 0.27% -8 0.07% 27 0.180 -0.11%

-7 -0.32% 27 -0.775 -0.05% -7 0.20% 27 0.490 0.09%

-6 -0.27% 27 -0.659 -0.32% -6 0.65% 27 1.587 0.74%

-5 -0.33% 27 -0.793 -0.64% -5 0.60% 27 1.448 1.34%

-4 0.62% 27 1.509 -0.02% -4 1.05% 27 2.560 *** 2.39%

-3 -0.07% 27 -0.174 -0.10% -3 2.11% 27 5.115 *** 4.50%

-2 0.52% 27 1.268 0.43% -2 1.96% 27 4.755 *** 6.46%

-1 -0.08% 27 -0.183 0.35% -1 3.16% 27 7.681 *** 9.62%

AD 0.86% 27 2.080 ** 1.21% CD -2.36% 27 -5.735 *** 7.26%

1 5.09% 27 12.371 *** 6.30% 1 -1.16% 27 -2.824 *** 6.10%

2 0.95% 27 2.303 ** 7.25% 2 -0.04% 27 -0.105 6.05%

3 1.55% 27 3.774 *** 8.80% 3 -0.34% 27 -0.834 5.71%

4 -0.33% 27 -0.802 8.47% 4 -0.63% 27 -1.531 5.08%

5 0.23% 27 0.556 8.70% 5 -0.58% 27 -1.405 4.50%

6 -0.94% 27 -2.282 ** 7.76% 6 0.06% 27 0.138 4.56%

7 -0.16% 27 -0.397 7.60% 7 -0.14% 27 -0.345 4.41%

8 -0.40% 27 -0.966 7.20% 8 -0.57% 27 -1.391 3.84%

9 -0.96% 27 -2.323 *** 6.24% 9 0.21% 27 0.498 4.05%

10 0.17% 27 0.412 6.41% 10 0.11% 27 0.267 4.16%

Table 3: Abnormal Returns for Additions and Deletions Surrounding AD and CD.Annoucement Day Basis Change Day Basis

(Note) 1. ***,**,* indicates that the null hypothesis of abnormal return is zero is rejected at 1%,5%, 10% level respectively.T-values are omitted. 2. MCAR is the cummulative MAR from AD-10 onwards.

12


Figure 1 describes MAR and MCAR around the AD and Figure 2 displays them around the CD. Figure 1-a indicates

the deletion portfolio’s behavior around the AD. There seems to be a steady decline before the AD but the degree is not

statistically significant. There is a possibility that some investors are trading on speculation or information leakage with

respect to their deletions, but they are affecting stock price only to a limited degree. On days after AD, the deletion

portfolio’s return dives. The decline continues over the sample period, and MCAR at AD+10 is -15.51%. Figure 1-b

indicates the addition portfolio’s behavior. Like the case with deletions, stocks tend to edge up before the AD but not

significantly. Return on the additions portfolio shoots up on days after AD and MCAR at AD+10 is +6.41%.

(Note) 1. MAR is the mean abnormal return for a given day computed from the cross sectional average of the samples. 2.MCAR is the mean cummulative abnormal returns. MCAR is the cummulated MAR from AD-10 onwards. 3. Actualannoucements becomes public knowledge after the market close on AD. Therefore the MAR is the largest on AD+1.

Figure1-b: Abnormal Returns for Additions Surrounding Annoucement Date

-2.00%

-1.00%

0.00%

1.00%

2.00%

3.00%

4.00%

5.00%

6.00%

-10 -8 -6 -4 -2 AD 2 4 6 8 10

Days Relative to Annoucement Date

MA

R

-2.00%

0.00%

2.00%

4.00%

6.00%

8.00%

10.00%

MC

AR

MAR MCAR

Figure 1-a：Abnormal Returns for Deletions Surrounding Annoucemnet Date

-5.00%

-4.00%

-3.00%

-2.00%

-1.00%

0.00%

1.00%

-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 AD 1 2 3 4 5 6 7 8 9 10

Days Relative to Announcement Date

MA

R

-18.00%

-16.00%

-14.00%

-12.00%

-10.00%

-8.00%

-6.00%

-4.00%

-2.00%

0.00%

2.00%

MC

AR

MAR MCAR

13


Comparison of the absolute magnitude of the event makes it apparent that deletions are the most affected.

Figures 2-a and 2-b also uncover another difference between deletions and additions. Cumulative abnormal returns are

maintained in the case of the former while the latter gives them back after the CD. Beneish & Gardner (1995) found

that additions to the Dow Jones Industrial Average made little difference between pre and post event while deletions

had a significant negative effect. My findings with respect to the Nikkei 225 are similar in a sense that deletions are the

most affected.

(Note) 1. MAR is the mean abnormal return for a given day computed from the cross sectional average of the samples. 2. MCAR is themean cummulative abnormal returns. MCAR is the cummulated MAR from AD-10 onwards.

Figure 2-b: Abnormal Returns For Additions Surrounding Effective Date

-3.00%

-2.00%

-1.00%

0.00%

1.00%

2.00%

3.00%

4.00%

-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 CD 1 2 3 4 5 6 7 8 9 10

Days Relative to Effective Date

MA

R

-2.00%

0.00%

2.00%

4.00%

6.00%

8.00%

10.00%

12.00%

MC

AR

MAR MCAR

Figure 2-a: Abnormal Returs for Deletions Surrounding Effective Date

-5.00%

-4.00%

-3.00%

-2.00%

-1.00%

0.00%

1.00%

2.00%

3.00%

4.00%

5.00%

6.00%

-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 CD 1 2 3 4 5 6 7 8 9 10

Days Relative to Effective Date

MA

R

-14.00%

-12.00%

-10.00%

-8.00%

-6.00%

-4.00%

-2.00%

0.00%

2.00%

MC

AR

MAR MCAR

14


B. Existing hypotheses and interpretation

The findings are not consistent with the price pressure hypothesis which considers the increased value of the

additions is due to temporary price pressure from noise traders. As noise traders buy the additions upon announcement,

the market maker facilitates the trade by selling them at a premium. The premium is compensation for the market

maker to unwind the position as the demand shock fades. By the same token, a discount is necessary for the market

maker to absorb the negative demand shock upon deletion announcement. If price pressure is the cause of price action,

post-event abnormal return should be negative. The findings indicate that both deletions and additions in the post-event

period fail to return to where they stood before announcement. The results also indicate that the event drives

equilibrium to a new level.

The imperfect substitute hypothesis implies that the price of a stock is vulnerable to an external shock when no

perfect substitute is available. When price moves above fair value, rational investors would sell the stock concerned and

substitute it in their portfolio with another stock of similar fundamentals. If availability of an appropriate substitute is

limited, rational investors would leave overvalued stocks as they are. An addition with a positive demand shock would

see price rise above fair value and stay there since there is little room for arbitrage. Correspondingly, triggered by a

negative demand shock, deletions would permanently decrease in value. However, there is asymmetrical return

between the addition and deletion samples. The deletion samples decline in value more than the addition samples rise.

This is not perfectly explainable by the imperfect substitute hypothesis. If the demand shock is the only cause of price

movement, one should be able to observe a similar price pattern for both additions and deletions. Therefore,

asymmetric price movement may indicate some other cause behind price movement. One may argue that the market is

vulnerable to negative news. Securities analysts tend to generate positive reports. Thus, market participants are used to

positive news. In such circumstances, negative news attracts investor attention more so than positive news.

4. Trading Simulation

One of the major arguments of the efficient market hypothesis is that arbitrageurs in the market behave in

a way to achieve such efficiency. When stock price moves away from fundamental value, arbitrageurs come in

and take the opposite position, which eventually drives value to its equivalent. Trading strategy typically

employed by arbitrageurs is simulated below.

A. Two types of trading strategy

Upon announcement of composite change, arbitrageurs go long on the stocks newly adopted and short sell

deletions. For short sales, they go to the securities lending market and borrow the stocks. Since borrowing demand in

the securities lending market shoots up upon announcements, they have to act quickly. This long position in additions

15


and short position in deletions will be held until one day prior to the change date. Arbitrageurs intend to unwind the

position as close as possible to the closing of the change date. Arbitrageurs would face price risk if they were unable to

unwind the position by the closing bell. There is a possibility that the stocks they have do not close due to the imbalance

of trades.8 Some arbitrageurs start unwinding positions 10 to 15 minutes prior to the close.

Arbitrageurs generate excess return using the following two trading strategies (Table 4). The first is typically

employed by arbitrageurs who anticipate a continuous price rise in the post-announcement period. The rise may be

caused by irrational traders who simply follow the “bullish” trend. In practice, such an arbitrageur shorts deletions (long

on additions) upon announcement. He keeps the position until one day prior to the change date and squares it as close

as possible to the market close of that day. He may want to wait until just before the closing bell knowing that

index-tracking investors will come in and buy the additions (sell the deletions) near the close. The second strategy is

based on the conviction that irrational noise traders who do not have any good reason to buy additions except for rising

momentum (sell the deletions except for falling momentum) will reverse their positions once such momentum

dissipates. Therefore, the strategy is to short the additions (go long on the deletions) at the closing price on one day prior

to the change date. Subscripts a, b, and c differ depending on how many days the positions are held. Strategy-a dictates

unwinding at the close of the change date while strategy-b on CD+1 and strategy-c on CD+2.

▲ ▲Announcement +1 +2 +3 . . . . -1 Change Date +1 +2 +3

(AD) (CD)

Deletions S S S Sq

Additions L L L Sq

Deletions L Sq

Additions S Sq

Deletions L L Sq

Additions S S Sq

Deletions L L L Sq

Additions S S S Sq

========================　・・・・=========================================>t

(Note) "S" indicates the short position while "L" the long position. "Sq" shows the traders to liqudate whatever position hecarried upto that day.

========================　・・・・=========================================>t

Table 4: Arbitrageurs' Trading Simulation

Strategy 2-a

Strategy 2-b

Strategy 2-c

Strategy 1

B. Simulation results

Table 5 describes average return of the strategies regarding 77 deletions and 69 additions. The same analysis was 8 When there is an order imbalance at the close, the Tokyo Stock Exchange uses a tentative closing price when the imbalance of trades is likely to create a gap in the closing price. The tentative price is calculated as the last traded price plus the predetermined price range.

16


also conducted based on the portfolio approach. Since the composite change event which occurred in April 2000 is

dominant in size, controlling it using the portfolio approach is useful.

Deletion Samples (n) 77 46 46 46

Average 10.61% 6.20% 7.41% 6.19%

Max 36.17% 26.11% 26.09% 26.96%

Min -9.87% -5.07% -10.87% -14.29%

% of winning trades 87.01% 86.96% 86.96% 84.78%

Deletion Events (n) 26 7 7 7

Average 7.97% 5.15% 7.13% 6.52%

Max 20.69% 10.16% 19.90% 19.05%

Min -7.93% 1.69% 1.21% -3.62%


Addition Samples (n) 69 69 69 69

Average 12.29% 3.34% 4.22% 4.52%

Max 46.97% 12.69% 15.99% 18.78%

Min -10.87% -6.09% -3.70% -11.19%


Deletion Events (n) 24 24 24 24

Average 6.17% 2.59% 3.50% 3.56%

Max 29.04% 8.05% 10.90% 12.38%

Min -10.87% -3.45% -0.80% -11.19%


Total Samples (n) 146 115 115 115

Average 11.27% 3.73% 4.53% 4.09%

Max 83.14% 30.13% 33.10% 33.26%

Min -8.99% -5.96% -8.17% -14.29%


Total Events (n) 28 28 28 28

Average 12.69% 3.51% 4.78% 4.68%

Max 36.62% 12.79% 23.79% 22.21%

Min -5.06% -1.66% -0.80% -11.19%


(Note) 1.Transaction costs such as brokerage comissions and securities lending fees areassumed to be zero. 2. Trading strategies on events are based on the equal weightedportfolio approach. Regardless of the size or number of the sample firms in a given event,traders are assumed to have invested equally.

Total

Table 5 Results of the Trading SimulationsStrategies employed for Deletions

Strategy 1 Strategy 2-a Strategy 2-b Strategy 2-c

Strategies employed for Additions

17


For deletions and additions, strategy-1 generates 12.69% return per event over the 12-year period. Among the

three sub-tactics of strategy-2, on the other hand, strategy 2-b seems to have generated the best performance both in the

individual stock approach and the portfolio approach. Selling additions at the close of CD-1 performed worse than

buying deletions at the close of CD-1 (strategy 2-b with 7.41% vs 4.22% in the individual stock approach, and 7.13%

vs 3.50% in the portfolio approach). This could mean that deletions may deviate more from fundamental values than

additions. If so, why do deletions perform worse? One may possibly hypothesize that that there is a barrier for

arbitrageurs to participate in deletions. Since arbitrageurs have to borrow deleted stocks from the securities lending

market during the event period, that may thwart non-institutional arbitrageurs from doing so. Thus, irrational

momentum traders may cause stocks to deviate more from equilibrium.

Finally, the winning percentage of strategies is computed. All of the pre-determined systematic strategies

generated profits 80% to 100% of the time. Considering the maximum and the minimum profit in the sample period,

the strategies are evaluated as extremely profitable. Figures 3-a and 3-b respectively depict performance of deletions

and additions over the 12-year period. Figure 3-c combines the two. Witness the continuously lucrative results from

event No. 1 through No. 39. The composite change event is a recurrent one. Therefore, one can easily predict that the

same trading pattern will repeat itself. If rational traders in the market act in a way that the theory suggests, there should

remain no arbitrage opportunities over such a long period of time. The fact that the recurrent events generate a similar

pattern of price predictability casts some doubt on the arbitrage argument in achieving the efficient market hypothesis.

18


(Note) 1. The white bar chart shows the results of the strategy 1 while the black bar indicates the returns from strategy 2-b. 2.Graph shows only the strategy 2-b for the simplicity reason. 3. Returns are not adjusted by the market model. Raw returns fromthe strategy are shown in all the three cases.

Figure 3-a: Trading Profits from Deletions

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

25.00%

1 2 3 4 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 23 25 28 29 31 32 36 37 39

Event No.

Ret

urn

Figure 3-b: Trading Profits from Additions

-15.00%

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

25.00%

30.00%

35.00%

1 2 3 4 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 23 25 28 29 31 32 36 37 39

Event No.

Ret

urn

Figure 3-c: Trading Profits from Total Index Change

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

25.00%

30.00%

35.00%

40.00%

1 2 3 4 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 23 25 28 29 31 32 36 37 39

Event No.

Ret

urn

19


C. Alternative hypotheses

Most prior studies try to explain abnormal returns assuming that arbitrage activity would drive a stock price to

its fundamental value. The very reason abnormal return is present is due to the cost of arbitrage. This logic is consistent

with the efficient market hypothesis where the market can eventually be efficient if it becomes frictionless. The

behavioral economists’ point of view, however, is different. Human beings sometimes behave in an irrational manner

and so does the market. Investors’ expected utility function is not monotonous as economists assume. They are risk

averse when obtaining profit but become risk takers once their wealth falls into the red. This kink in the investors’ utility

function was originally pointed out by Tversky and Kahneman (1979) in their well-known prospect theory.

Since the advent of behavioral finance, there are a number of theories that describe the situation where investors

behave in an irrational manner. In the case of index change, the positive feedback traders in the market may have played

an important role by taking a position in the same direction as the event suggests. For additions, they go long on the

stocks concerned and go short on deletions. Then some rational traders accelerate the momentum by also buying the

additions and selling deletions. As De Long et al. (1990) suggest in their theory, rational investors do not necessarily

counter trade with noise traders when they anticipate positive feedback traders will further push up the price. This

theory is compatible with the price behavior in the index change. The initial upward drift may be caused by some index

tracking investors who have no choice but to buy a certain number of shares by the change date. The upward drift

catches the positive feedback traders’ attention and they buy the additions as long as they go up. Rational traders will

also buy the stock (although it deviates from its fundamental value) since they know positive feedback traders will pay

an even higher price as long as upward momentum is intact. Such excess demand will only be met at a higher price.

The reverse is true for deletions. Upon announcement, index tracking investors dump the stocks, causing the initial

negative drift, which will subsequently trigger positive feedback traders’ sales. Since no rational investor would counter

such trading until the event is over, negative abnormal return for deletions is detected in the event period.

Does the index change have some features that facilitate such trading patterns? One has to be aware of the two

psychological traits that make this event special. Firstly, investors tend to react to news that has short-term

consequences rather than long-term ones as pointed out by Froot et al. (1992). Index change indeed has an impact on

short-term consequences. Therefore, there is good chance that many investors participate in the event. Secondly, as

Schafstein and Stein (1990) argue, institutional investors (index-fund managers) are penalized should they make a bad

decision alone. However, if their colleagues make the same bad decision, they will not be penalized. This asymmetric

compensation system induces them to take a similar action for an uncertain event such as index change. In this

case, they choose to take no action until right before the change date to minimize tracking error. Despite the fact

that selling deletions and buying additions is likely to generate higher expected return for a given risk, institutional

20


investors concentrate trading in the vicinity of the close on CD-1. This is part of the reason why index tracking

investors are willing to pay an overvalued price for additions and sell deletions at an undervalued price.

Conclusion

This empirical study of Nikkei 225 constituent changes sheds light on the market price of stocks from the

following three perspectives. Despite the fact that there is no obvious reason to believe, ex-ante, that index change

contains no new information, additions to the Nikkei 225 demonstrate statistically significant positive abnormal returns

and deletions negative abnormal returns. This is inconsistent with Scholes (1970)’s argument where stocks do not react

to non-information. Secondly, abnormal returns both on the additions and deletions are observed not only on the

announcement date but also on other days in the post-announcement period. This may suggest that positive feedback

traders exist behind such price action. Lastly, and most interestingly, despite the recurrent nature of index change, price

predictability is evidenced. Additions rise until CD-1 and some of the gain is reversed in the post-CD period, but not all.

Deletions decline until CD-1 and regain some of the loss in the post-CD period, but not all. This pattern by and large

remains the same for the 12-year period.

The conclusion here is that mispricing in the market may not disappear even if a frictionless environment is

achieved. The evidence presented here is contradictory to the efficient market hypothesis.

Here, I would like to express my gratitude to Prof. Shigeki Sakakibara, Prof. Nobuyuki Isagawa, and Prof. Yoshitaka

Kai (Graduate School of Business Administration, KOBE UNIVERSITY) for their useful advice, as well as the

referees for their valuable comments.

(This paper was submitted, selected and published as it was.)

2


References

Beneish, D.M. and Gardner C.J. （1995）, “Information Costs and Liquidity Effects from Changes in the Dow

Jones Industrial Average List”, Journal of Financial and Quantitative Analysis Vol. 30 No. 1, pp135-157

De Long, J. B., Shleifer, A., Summers, H.L. and Waldman, J.R. （1990）, “Positive Feedback Investment

Strategies and Destabilizing Rational Speculation”, Journal of Finance 45, pp379-395

Dhillon, U. and Johnson, H. （1991）, “Changes in the Standard and Poor’s 500 List” Journal of Business Vol.

64 No. 1, pp75-85

Froot, Kenneth A., Scharfstein, S. D. and Stein, C. J. （1992）, “Herd on the street: Informational inefficiencies

in a market with short term speculation”, Journal of Finance 47, pp1461-1484

Hanaeda, H. and Serita, T, （2003）, “Price and Volume Effects Associated with a Change in the Nikkei 225

List: New Evidence from the Big Change on April 2000” International Finance Review 4, pp199-225

Harris, L., and Gurel, E. （1986）, “Price and Volume Effects Associated with Changes in the S&P500 List:

New Evidence for the Existence of Price Pressure” Journal of Finance 41, pp815-829

Jain, P. （1987）, “The Effect on Stock Price from inclusion in or exclusion from the S&P500”, Financial

Analysts Journal 43, pp58-65

Kahneman, D and A. Tversky (1979), “Prospect Theory”, Econometrica

Lynch, W.A. and Mendenhall, R.R. （1997）, “New Evidence on Stock Price Effects Associated with Changes

in the S&P 500 Index”, Journal of Business Vol. 70 No. 3, pp351-383

Liu S. （2000）, “Changes in the Nikkei 500: New Evidence for Downward Sloping Demand Curves for

Stocks”, International Review of Finance, 1:4, pp245-267

Saito, M and M. Onishi（2001）, “Nikkei Index Change – Its Impact on Stocks’ Liquidity”, Modern Finance No.

9 pp67-82 (in Japanese)

Scholes, S. M. （1972）, “Market for Securities: Substitution versus Price Pressure and the Effects of

Information on Share Prices” Journal of Business 45, pp179-211.

Sharfstein, David S., and Jeremy C. Stein （1990）, “Herd behavior and investment” American Economic

Review 80, pp465-479

Shleifer, Andrei （2000） “Inefficient Markets”, Oxford University Press

Serita, T.（1996）, “Index transaction and stock prices – Event Study Approach”, Aoyama Economic Review No.

48 Vol. 1, pp 61-80 (in Japanese)

Watakori, N（2001）, “Downward shift of the Nikkei 225 - Nikkei Index Change” Securities Analyst Journal

2242, pp109-134 (in Japanese)

Wurgler, J. and Zhuravskaya, E. （2002）, “Does Arbitrage Flatten Demand Curves for Stocks?” Journal of

Business Vol. 75 No. 4, pp583-608

Changes in Nikkei 225 Composite Stocks- Trading Simulation ... · stocks (deletions) from the...

Documents

Transcript of Changes in Nikkei 225 Composite Stocks- Trading Simulation ... · stocks (deletions) from the...